NextFin News - Can artificial intelligence lift emerging economies out of a structural growth slowdown, or will it widen the gap between countries that can deploy it and those that cannot? The World Bank’s June 2026 outlook puts global growth at 2.5% this year, down from 2.9% in 2025, while its longer-run estimates show potential growth falling to about 2.2% in the 2020s from 3.6% in the 2000s. The Bank’s message is therefore more demanding than “use AI”: adoption is an opportunity, but the growth dividend depends on power, connectivity, data, skills and institutions.
The timing matters. The World Bank says emerging market and developing economies face their weakest per-capita income growth since the pandemic, just as productivity has become the binding constraint on expansion. Its Global Economic Prospects report identifies broader AI adoption as an upside risk, but also says countries that cannot adopt AI widely risk falling further behind. That creates a policy race with an unusual feature: the technology can be bought, but the economic capacity to use it cannot be imported at the same speed.
For investors and policymakers, the implication is not that every emerging market should build a frontier model. It is that the next phase of AI’s global economic effect will be decided by diffusion: whether firms, governments and workers can adapt existing tools to local production, finance, education and health. Countries that make AI a general productivity layer have a better chance of reversing weak potential growth than those that simply announce national AI strategies.
The Growth Problem Is Structural Before It Is Technological
The World Bank’s figures describe a regime change in the growth backdrop, not a routine pause. Global potential growth, the maximum sustainable pace that does not generate inflationary pressure, fell from 3.6% a year in the 2000s to 2.8% in the 2010s and is projected at about 2.2% in the 2020s. In emerging market and developing economies, the decline is steeper: potential growth is expected to fall from 5.9% in 2000–09 to 4.1% in 2020–29.
The composition of the slowdown matters as much as the headline. The Bank says weaker total-factor-productivity growth accounted for more than 40% of the decline in EMDE potential growth. Global TFP growth fell from about 1.4% annually in the 2000s to 1.1% in the 2010s and is projected to average about 0.8% in the 2020s. Investment and labor-force expansion have also weakened. That combination cannot be repaired by a single interest-rate cycle or a temporary rebound in commodity demand.
Near-term conditions reinforce the pressure. The Bank projects global growth at 2.5% in 2026, down from 2.9% in 2025, with energy-market disruption and renewed inflation adding to borrowing costs and fiscal strain. The headline is cyclical in part: energy supplies can recover, trade can strengthen and regional activity can rebound. But the potential-growth trend is structural. Even if the current shock fades, economies return to a lower speed limit unless productivity, capital formation or labor supply changes.
That is why AI enters the discussion as more than a technology theme. If it raises output per worker, improves the allocation of capital and reduces the cost of delivering public services, it could lift the economy’s supply capacity rather than merely pull demand forward. But if its benefits remain concentrated in a few large firms or in advanced economies with abundant computing power, it will not reverse the EMDE slowdown. It will make the divergence more visible.
AI’s First Channel Is Investment; Its Real Test Is Diffusion
The immediate AI growth story is easier to observe in advanced economies than in emerging ones. The World Bank’s July 2026 analysis says AI is likely to lift growth in the near term mainly through investment, especially in advanced economies. Spending on data centers, chips, cloud capacity and software raises demand before it proves that those assets have produced economy-wide efficiency.
For emerging economies, the more important channel is adoption of existing tools. A bank that uses machine learning to assess borrowers with limited conventional credit histories can expand access to finance. A small business that uses an AI advisory tool through a smartphone can improve pricing, inventory and customer service. A health system can use decision-support tools to stretch scarce clinical capacity, while an education system can tailor instruction where teachers are in short supply. These are not frontier-model achievements. They are organizational changes that convert computing into output.
“There is optimism about AI enabling developing countries to leapfrog development challenges by reducing human error, optimizing complex production and distribution processes, and facilitating decision-making.” — World Bank, World Development Report 2026
The mechanism runs through complementary inputs. The World Bank’s Digital Progress and Trends Report 2025 organizes them as four Cs: connectivity, compute, context and competency. Connectivity means reliable energy and digital networks. Compute covers chips, data centers and cloud services. Context means usable, governed data. Competency means workers and managers who can apply the tools.
Each missing input reduces the return on the others. Faster networks do little for a firm without data or management capacity. A data center without reliable electricity becomes an expensive bottleneck. Skilled workers cannot deploy models where data rules are unclear or cloud access is unaffordable. This is why AI adoption resembles a production system rather than a software download: the weakest complement constrains the aggregate gain.
The capital-allocation implication is consequently broader than semiconductor demand. Power generation, transmission, broadband, cloud access, cybersecurity, digital identity and payments can all become part of the AI diffusion stack. Yet the distribution of returns will depend on competition. If local businesses can only rent expensive services from a concentrated group of foreign providers, some productivity gains can leak abroad through licensing, cloud and data costs.
The first-order effect is more AI-related capital spending. The second-order effect is a reallocation of development capital toward the infrastructure that lets ordinary firms use AI. The third-order question is whether that spending changes measured productivity or merely produces a new layer of imported technology. The latter outcome can support equipment suppliers while leaving domestic value added and wages largely unchanged.
Adoption, Not Frontier Research, Is the Emerging-Market Fault Line
The World Bank’s evidence points to a structural distinction between developing countries that adopt and adapt AI and those that remain dependent on technology produced elsewhere. Frontier research requires scarce talent, enormous compute and deep pools of data. Adoption can begin with commercially available models, but adaptation is needed to make them useful in local languages, industries and regulatory settings.
That distinction changes the policy test. A country does not need to duplicate the largest model developers to improve agricultural extension, tax collection, logistics or small-business finance. It does need procurement rules that allow public agencies to buy useful tools, data governance that permits responsible reuse, digital infrastructure that reaches firms, and training that gives workers a path into complementary tasks.
The 2026 World Development Report says AI can fill skills gaps in education and health and make small enterprises more productive through smartphones and high-speed internet. It also identifies the risk that AI will widen the gap between higher- and lower-income countries because of its requirements for computing power, data and skills. The same technology therefore carries two different development functions: it can augment workers where institutions diffuse it, or automate tasks and lower wages where workers have little bargaining power or access to retraining.
“Such optimism should not be unbridled.” — World Bank, World Development Report 2026
That warning is central to the labor-market mechanism. In economies where many workers and small firms have limited access to advanced tools, AI may raise productivity by making advice and specialized capabilities more widely available. But in export-oriented services that rely on routine digital tasks, generative AI could reduce the advantage of low-cost labor. The same country could gain from productivity in agriculture and lose jobs in business-process outsourcing.
The cross-industry effect is also asymmetric. Firms that own customer data, distribution networks or trusted brands can combine AI with existing assets. Smaller firms may gain access to the same models but lack the capital to reorganize workflows. That can increase concentration even when the technology is widely available. Competition policy is not an accessory to AI strategy; it helps determine whether adoption broadens the productive base.
For markets, this shifts attention from a narrow model-building story toward a layered infrastructure and applications story. The sectors exposed to the diffusion cycle include power, connectivity, cloud and enterprise integration, but their long-run demand depends on whether local firms and public agencies adopt the services at scale. Routine digital service exporters and workers whose tasks are easy to automate face greater adjustment risk if complementary job creation does not follow.
The Counter-Thesis: AI May Reinforce the Existing Divide
The strongest argument against the World Bank’s AI-upside case is not that the technology lacks power. It is that the economics of power, data and talent may concentrate the gains in countries that already possess them. High-income technology companies can finance frontier research, train large models and deploy them across global markets. Emerging economies may pay for access while absorbing the adjustment costs.
This counter-thesis attacks the foundation of the adoption story. If AI substitutes for routine labor faster than it complements workers, countries that built development strategies around low-cost digital services could lose export competitiveness. If models work poorly in local languages or depend on data that governments cannot safely share, the promise of adaptation remains theoretical. If electricity and broadband remain unreliable, the firms most in need of productivity tools will be least able to use them.
The World Bank itself identifies these risks: AI could widen the income gap, automate tasks that result in job losses or lower wages, and strengthen the advantage of a few technology companies based in high-income countries. The institution also warns that flawed or biased systems can produce worse decisions when safeguards are weak. Those are not edge cases. They are the channels through which a productivity technology becomes a distributional shock.
The answer is not to dismiss the technology, but to narrow the claim. AI is not an automatic catch-up machine. It is a force multiplier. It multiplies the quality of a country’s electricity, data, skills, management and institutions. Where those foundations are weak, it multiplies dependence and concentration instead.
The clearest falsifying signal for the positive thesis would be a sustained failure of AI-intensive productivity to spread beyond advanced economies: if, through 2028, productivity gains and AI-enabled investment remain concentrated in high-income countries while EMDE TFP growth stays near or below the World Bank’s roughly 0.8% global TFP baseline and the gap in potential growth continues widening, the claim that diffusion is an emerging-market growth engine would be wrong. That is a measurable test, not a mood.
What the World Bank’s Message Means for Markets
The report’s market channel is thematic rather than a verified same-day repricing of emerging-market assets. AI-related investment can support demand for data-center equipment, cloud services, power infrastructure and digital networks. But the near-term spending impulse should be strongest where financing, grid capacity and enterprise demand already exist, which tends to favor advanced economies and a limited set of emerging-market hubs.
Over the medium term, the opportunity shifts toward countries that can convert imported AI into domestic productivity. The relevant indicators are not model size or the number of AI announcements. They are reliable electricity, broadband penetration, cloud affordability, digital-payment usage, data quality, worker training and measurable productivity in sectors such as logistics, agriculture, finance and public administration. A country that improves those complements can capture more value from each dollar of imported compute.
Over the long term, the distribution of gains can shape currencies, sovereign credit and equity-sector leadership through productivity and fiscal capacity. Faster productivity can raise potential output, improve debt sustainability and support wages without the same inflation pressure. But a technology-importing economy with weak domestic value capture could face higher external payments for software and cloud services without a comparable improvement in its trade balance.
The base case is conditional diffusion. AI investment continues to support global activity, while EMDE productivity gains arrive unevenly and concentrate in countries with credible digital and energy build-outs. The upside case is a broad adoption wave in which low-cost, general-purpose tools reach small firms and public services, pushing productivity growth above the World Bank’s baseline and narrowing selected development gaps. Its trigger would be repeated evidence of productivity gains outside frontier technology sectors.
The downside case is a dual-speed AI economy. Advanced economies absorb the investment and innovation gains; emerging economies face job displacement, higher technology-import bills and a wider skills gap. Its trigger would be falling employment or wages in routine digital services without offsetting gains in higher-productivity sectors, alongside continued infrastructure shortfalls.
None of these scenarios is captured by a single AI spending number. The technology is a structural potential offset to a structural growth slowdown, but the offset is not portable by itself. Policy determines whether the productivity function moves outward or whether the country simply rents access to someone else’s platform.
The World Bank’s call should therefore be read as a development-infrastructure agenda with an AI label. In the short term, sentiment may favor visible technology investment. In the medium term, fundamentals will favor countries and firms that demonstrate adoption and productivity. In the long term, the dividing line will be institutional: who can turn power, data and skills into broad-based output rather than narrow platform dependence.
AI will not rescue emerging economies merely because the models improve. It will lift growth where the complementary systems are built fast enough to let ordinary firms and workers use them. The next AI divide is therefore likely to be measured less by who invents the model than by who captures the productivity.
Data cutoff: August 4, 2026, 13:17 UTC.
Explore more exclusive insights at nextfin.ai.
